Nanopore sequencing and multiomics reveal predictable non-coding RNA activation in DNA methylation deficient Arabidopsis thaliana

Long intergenic non-coding RNAs (lincRNAs) are regulatory transcripts from intergenic regions with diverse expression patterns, but whether the activation of lincRNA is widespread and predictable remains unclear. Here, we applied Oxford Nanopore Technology Direct RNA and DNA sequencing (ONT DRS and DDS) to Arabidopsis DNA methylation-deficient mutants (ddm1 and met1) and wild type. Differential expression analysis identified 340 upregulated lincRNAs and 209 lincRNAs with consistent expression whose expression was negatively correlated with DNA methylation. Similar activation patterns were also detected in natural populations. To further characterize these lincRNAs, fifty multi-omics features were compiled to train six machine learning models for classifying ddm1-activated lincRNAs and Random Forest achieved the highest average precision of 0.96. Feature importance analysis highlighted population-level DNA methylation, ONT-derived RNA modification and transposable elements as key predictors. These results indicate that epigenetic variation shapes predictable lincRNA activation, establishing a framework for systematic discovery of expressible non-coding RNAs.